Evaluation of the risk of death for very low-birthweight babies and comparation between neonatal intensive care units: Application of ROC curves

In general, the neonatal rates of mortality are considered the most important measure for the evaluation of health services and the development of society. Babies with very low birthweight (<1500g at birth) contribute significantly to the increase of the mortality and morbidity rates. This study aims to compare, in the same set of babies, 4 different indexes of clinical seriousness; these indexes quantify the risk of death for babies with very low birthweight. The study was conducted on 169 new-borns with birthweight under 1500g, admitted to the Neonatology Unit of the Hospital Garcia de Orta in Portugal, between January 1992 and July 1995. The data were collected retrospectively on the same set of new-horns to permit the comparison of the different indexes. To compare Neonatal Intensive Care Units, we use another set of new-borns. These data (234 new-borns with very low birthweight) were collected in 4 different Portuguese hospitals (H-1, H-2, H-3 and H-4) during 1995. The 4 indexes studied were: CRIB (Clinical Risk Index for Babies), TISS (Neonatal Therapeutical Intervention Score System), SNAP (Score for Neonatal Acute Physiology) and SNAP-PE (Score for Neonatal Acute Physiology-Perinatal Extension). Also, included in the study was the variable WEIGHT (birthweight) because it is an important index to evaluate the initial risk of death. The selection of the best diagnostic test was based on the Receiver Operating Characteristic (RDC) curves. The results, based on the ROC curves, allow the conclusion that CRIB is the best index for evaluating the risk of neonatal death, and that hospital H-4 has better performance in terms of neonatal intensive care.